Transfering Calls

Transfer reasons are categorized by the specific intelligence task involved.

Global Transfer Reasons

These apply across all intelligence tasks:

• The user is frustrated, upset or appears to be agitated, or uses ANY expletives.
• The user asks to be transferred to another representative, or just requests to be transferred in general.

Right Party Verification

Gather Identity Information

• If the user refuses to provide any piece of information.
• The user wants to speak to a different agent, to a different department, or to a supervisor, or to someone else, or to a human.

Gather Callback Number

• The user will not provide a phone number.

Gather Name/Address Retry

• If the user refuses to provide their name/address again.

Verification

If the caller fails verification, the call will be tranferred to a live rep. (see Right Party Verification document for criteria)

Multi Choice Prompt

• If the user refuses to provide an option.

Call Intent Handling

Below are the possible transfer reasons for when the AI is assisting the caller with their account balance.

• If the user wants to make a payment.
• If the user wants to set up a payment arrangement.
• If the user wants to add their balance to their current payment arrangement.
• If the user still has questions about their balance or about something else.
• If the user wants to check the account balance of their husband, wife, child, etc.
• If the user has any questions about their payment arrangement outside of information about their last or next payment.
• If the user wants an explanation of their balance/what the balance is for.
• If the user disputes the balance and is upset about it.

Arbitration

In some cases the AI models may disagree as to which course of action should occur next, this is called arbitration and always results in a forced transfer.

When Does Arbitration Happen?

Arbitrations can occur at any point during the conversation. but it is most commonly triggered when one model believes the system should proceed to a new state, while another model does not. Typically, this involves a conflict between the Action Model and the Conversation Model:

-The Action Model is responsible for determining when to take high-level actions, such as transferring to a human, executing a transition, or remaining within an intelligence task (no action). -The Conversation Model is focused on the flow of dialogue, such as asking questions, and gathering information or inputs from the user.

Example Scenario

• The Conversation Model has been collecting the caller's identifying information (e.g., name, date of birth, address).
• At some point, the Action Model determines that sufficient information has been gathered and it's time to transition to gathering the caller's intent.
• However, the Conversation Model continues prompting the user for information, thinking it has not achieved its goal.

This misalignment results in arbitration. Since the system's models are no longer in agreement and could lead to user confusion or a conversational loop, a forced transfer occurs to ensure continuity and avoid frustration.

Why Force A Transfer?

Transferring because of arbitration is an intentional safeguard designed to preserve reliability and prevent degraded user experiences.

Rather than risk:

-Repetitive or conflicting prompts,
-Redundant data collection,
-Misinterpretation of user intent,

…the system escalates to a live representative. This maintains a high standard of user experience and trust.